关键词: 变化检测, 合成孔径雷达（SAR）图像, 多智能体遗传算法

Abstract: As the conventional evolutionary algorithms are often easy to trap in local optimum value, slowly convergent and time-consuming in dealing with the problem of Synthetic Aperture Radar （SAR） images change detection. In order to solve the problems, this paper proposes an unsupervised technique, namely multiagent genetic algorithm. The processed images are used to construct the difference image by logarithmic ratio, and the difference image is processed with medium value filter. Then the method takes the gray value as input information, uses the multi-agent genetic algorithm to search the global threshold. According to the global threshold, it gets the results in change detection. Simulation results show that the proposed method is more accuracy classification, faster convergence rate and more efficient than GA, ICSA.

Key words: change detection, Synthetic Aperture Radar（SAR） image, multiagent genetic algorithm